Bayesian Process Errors
If I run a Bayesian process and get just the following errors, relating to WAIC and LOO, is the BIC value still likely to be reliable?
Warning: WAIC estimate unreliable - at least one effective parameter estimate (p_waic) larger than 0.4. We recommend using LOO instead.
Warning: LOO estimate unreliable - at least one observation with shape parameter (k) of the generalized Pareto distribution higher than 0.5.
Comments
I do not immediately see how the problems with WAIC and LOO would imply problems with BIC. The BIC might be "unreliable" if the MLE is borderline identified.
EJ
@EJ RE: "if the MLE is borderline identified"
Is there any diagnostic for this in the output of JASP? Or way that it can be inferred? Perhaps by SE in the MLE, if there is any way to see this?
If there is no hard error, this can be tricky to diagnose, in any program.